An Improved Wavelength-based Distributed Underwater Image Enhancement Algorithm
An improved wavelength-based distributed deep neural network model is proposed to address the attenuation and scattering of light during underwater propagation,which often leads to color distortion or low contrast in captured underwater images.The model resets the hyperparameter values to smooth out the gradient swing amplitude of algorithm optimization,and introduces a dropout layer to alleviate overfitting of the network model.Quantitative and qualitative experiments on publicly available dataset are conducted for the improved model and four existing underwater image enhancement algorithms.The experimental results show that the proposed model outperforms comparison algorithms in enhancing underwater images with different color distortions,and still has a good enhancement effect on underwater images with noise interference.
万方数据
维普
